# Full text of "An analysis of the modal split between airport limousine and taxicab for trips from the Chicago Loop to O'Hare International Airport and the implications for a possible future high-speed rail limousine service"

## See other formats

An Analysis of the Modal Split Between Airport Limousine and Taxicab for Trips from The Chicago Loop to O'Hare International Airport and the Implications for a Possible Future High-Speed Rail Limousine Service Master's Thesis By Michael G. Kennedy June, 1971 HE 9797.3 A532 3 5556 021 099 767 ■NORTHWESTERN UNIVER SI TY AN ANALYSIS OP THE MODAL SPLIT BETWEEN AIRPORT LIMOUSINE AND TAXI CAB TOR TRIPS PROM rpTTi? nifTOAftO T 0015 mo r> t "v-TAutr ttttte?t>'MA T»TO"MAT * rpwoni ■AND THE IMPLICATIONS J OR A POSSIBLE FUTURE HIGH- SPEED RAIL LIMOUSINE SERVICE A THESIS SUBMITTED TO THE DEPARTMENT OP CIVIL ENGINEERING IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE MASTER OP SCIENCE FIELD OF CIVIL ENGINEERING TRANSPORTATION LIBRARY MAY by Micnael G. Kennedy Evan st on , I llinoi s June s 1971 NORTHWESTER! ' . UBR»RV ?7?73 Pi £"32- ABSTRACT The subject of this analysis is the development of a theoretical node!* for predicting modal split between airport limousine and taxicab for trips from the Chicago Loop to 0'IIare Airport and to determine if this model might be useful for evaluating a future "rail limousine'' service. The input data for the independent variables, which consisted of travel times, costs, frequencies, and access distances for each mode, were gathered in the field by driving existing limousine and taxi- cab routes to O'Hare Airport. The best model was one which ex- pressed each variable as a difference between each mode. The resulting model was fo M.S. (mode split for limousine) ~ 86.7 - 1.5TT - 43. 7W, where "TT" is the* travel time difference (limousine -taxi) measured in minutes and "W" is tiie access distance difference ( limousine -taxi) measured in miles. Cost was found to be in- significant in the mode choice decision. At the outset of this study, it had been intended to. develop a practical model. How- ever, for reasons beyond the author's control, the critical dependent variable data, mode split by zone, from a public survy become "unavailable" . A hypothesized set of zonal mode * That is, a model not based entirely on fact. Part of the model's data base was hypothesized, consequently the resulting model was considered as "theoretically . # "splits was used instead of the survey modal splits. A rail- limousine line to O'Kare from one terminal in the Loop using the C.&M.W.R.R, right-of -way instead of the CTA right-of-way from the Loop to Jefferson Lark was found to be only slightly better than present bus limousine service at non-rush hours. I ACKITOV/LSPGEl^NTS The author wishes to acknowledge the assistance given by the following in the preparation of this thesis: Mr . 1 . 1 ' e rn e 1 a ( C h i c a go Ar e a T r an s p or t a t ion Study), Mr . T . Lisco (Chicago Area Transportation Study) , and especially Dr. P. Stopher who was my adviser. IV TABLE OP CONTENTS PAGE Abstract ii Acknowledgements iv List of Figures vi List of Tables vii I Introduction 1 II Airport Access Worldwide 3 III Ground Access to O'Hare International Airport 7 IV Proposed Model and Data Gathering 9 V Model Development and Analysis 21 VI Model Results 30 VII Implications for a Possible Future High Speed Rail Limousine Service to O'Hare Airport 38 VIII Conclusions 43 Bibliography 45 Appendix 49 LIST OP FIGURES FIGURE 1 2 Ciiicage Central Business District Arialy si s Zones PAGE 12 13 VI LIST OF TABLES TABLE PAGE I Zonal Centroids 16 II Data Base 18 III Summary of Model Development 23 IV Equations Resulting from Random Change 32 V Sensitivity Analysis 36 .vi i ' 2 HAPTER I - INTRODUCTION The subject of this study is the analysis of the present modal split between two nodes of transportation en- ployed for a very special urban trip. This journey, which represents less than one percent of all urban trips,! is from a central business district to a metropolitan airport and is made by air travelers. This study is one of the first times that an analytical model has ever been constructed to predict modal split for this quite interesting mode choice. The specific case examined was that of trips from Chicago's Loop to OMIare International Airport by airport limousine and taxi. As a helpful addition to this introduction, a brief outline of this work will now be given* In the next two chap- ters, the problem of airport access is examined, first world- wide and then in Chicago. Chapter four presents various types of modal split models, which are examined. It also covers the data gathering stage of this thesis. In the next chapter, the modal split model is statistically developed by means of a multivariate linear regression analysis. In chapter seven, the model and its implications are than used to consider possible high-speed rail limousine service* from the Loop to 0TIiare Airport *A rail limousine service is simply an express transit service on rails, using equipment similar to either the newest CTA or C.&N.W.R.R. cars, and offering amenities comparable to the present bus limousine service. - 1 - - Finally, conclusions concerning the model and its implications are drawn in chapter eight. - 2 - • CHAt; TER „ II _- , AiJ^RT . AC C ^§ 3„I^0xLl)V:ID5 In ths United States in 1950, 17.5 million people traveled -by air, representing 15% of all intercity travel by 2 common carrier. In 1957, there were 119 million airline passengers - - representing 63% of all trips by public carrier - - an increase of over 500% in only 17 years, Thus, travel by air is now the predominant form of intercity common carrier transportation in the United States. The typical air trip consists of three parts. Two are the ground access portions to and from the airport, and the other is the actual air flight. Presently, from 22% to 65% of total travel time is spent on the ground. There are a number of travel modes which one can use for the airport access segment of the trip. The following are the most commonly used modes: private car, rented car, taxicab, limousine, transit bus, rail rapid transit, railroad, or helicopter. Of these available modes, bhe predominant form for airport access transportation in the United States is the private automobile, which is employed by 37% to 74% of all airline passengers. Taxis account for from 12% to 41%, and limousines from 18% to 38%4of the access mode marke t . Ground access to most airports is a very slow process. Peak travel speeds to the major airports from the central busi- ness districts range from only 9.6 mph in Boston (13% freeway) and 23.4 mph in Chicago (85$ freeway) bo a high of only 35 mph in St. Louis (93$ freeway). The cause of the slow driving speeds is a situation which is common to all large cities, both in the United States and abroad - - traffic congestion. The causes of this congestion on our urban streets end freeways are many, The population explosion along with the affluent society, has caused automobile registration and subsequent auto usage in the United States to skyrocket in the last 20 years. Urban sprawl and its diverse origin-destination pattern, coupled with a lack of attention towards providing adequate or any riass tran- sit are but some of the many factors which contribute to the omnipresent traffic congestion. To further complicate the situ- ation it is anticipated that traffic congestion will continue in the future. As a result, driving tines to airports will probably increase, thus causing the airline passenger to spend an even greater portion of his total travel tine on the ground. Not only will the time spent getting to the airport increase, but congestion at most airports will increase. With the age of the jumbo jet now upon us, the Federal Aviation Administration has predicted a 440$ increase in air travel from 19(35 to 1980. »™ Even by the year 1975, the number of air travelers will equal the entire population of the United States in 1955. Also, so many more people can arrive at one tine now in a jumbo jet. The Boeing 747' s can carry more than 260$ g capacity of the present 707' s. Therefore, more people will demand ground transport and parking at the same time. All these people simply cannot be accommodated at the larger air- ports, For example, John F. Kennedy Airport in New York City is expected to handle 25,056 passengers A in 19G0 in the peak hour. These people combined with the people meeting them would require, if all were to use auto, an expressway at least 14 lanes wide.* Confronted by these statistics, it is questionable whether an\"all highway" solution to the airport ground access problem will provide for a viable situation. Some complementary means of transport must be examined. To date, four cities in the world have supplementary non-highway access to their airports. Gatwick Airport, located 23 miles south of London, was the first airport in the world to have rail and highway access in one coordinated passenger termin- al. British Rail provides express service to and from Victoria every 15 minutes 14 hours a day. This service is such that 69$ of all enplaning passengers use it. Brussels also has rail- ] 2 road service to its major airport." Tokyo has an 8.2 mile mono- 1 ^ rail line to its airport. The only city in the world to have a rail rapid transit line to its airport is Cleveland, Ohio. With the recent extension of its west Side rapid line, the *This is based on one car meeting each passenger and a capacity of 2,000 cars/hour for each expressway lane. J Cleveland Transit System provides frequent service, and a travel time of 22 minutes from Henry Hopkins Airport, to Down- tov/n Cleveland twenty-four hours a day, seven days a week. Presently two cities are in the planning stages of providing high speed rail service to their airports. The New York Board of Estimates has unanimously approved the construc- tion of a high speed railroad line from midtown Manhattan to Kennedy International Airport, although the New York State Legislature has stalled the project. Most of the right-of-way will be on present Long Island Railroad trackage. Operation was expected in late 1972, -5 but this seems quite doubtful now. London is now considering rail service to its largest airport, Heathrow. The British Airports Authority has entered Into discussions to consider establishing two rail links to Heathrow, one would be railroad service by British Rail and the other ] 6 would be rail rapid transit service by London Transport." These five cities are leading the way toward improved ground access to highly congested airports. -- 6 - .CHAPTER III - GROUND ACCESS TO O'EARE INTERNATIONAL AIRPORT O'Hare International Airport is located in the North- west corner of the city of Chicago about 18 miles from the Chicago Loop. Presently, driving time to O'Hare from the Loop 1 7 varies anywhere from 29 to 55 minutes, by time of day, ' on an average day. O'Hare Airport is the busiest in the world. During 1970, it is estimated that approximately =50,000,000 passengers will use OfEare Airport compared to only 25,000,000 at Kennedy International in Nov/ York. Furthermore, 12,000 passengers are expected to be handled during a typical peak hour at 0THare in 1970. On a weekday in October, 1964, the City of Chicago along with various co-operating agencies conducted a survey of 19 all airline passengers and crews.- The results of the survey showed that 60,954 passengers and airline crews used O'Hare that day, of which 48$ were through passengers. Of the other passengers, 45.9$ were from beyond the city of Chicago, 32.7$ were from the Loop and 21.4$ were from the rest of the city of Chicago. Furthermore, for all departing passengers from the Loop, over 75$ used limousine or taxi. Access to C'Kare from the Loop is provided by a spur of the only expressway to Northwest Chicago, the J.F. Kennedy Expressway. This expressway is six lanes wide and is presently \ 7 - } loaded beyond its capacity at rush hour, By 1980, the Federal Aviation Administration has predicted that 28,000 people will use 0*Hare during a peak hour. If half of these people are as- sumed to be through passengers, then 14,000 people will use the aforementioned six lane expressway, at least for a short distance. This many people would require a seven lane expressway for air- line passengers alone, if all arrived by automobile, Clearly some other form of ground access must be forthcoming. From the previous discussion it is obvious that o traffic congestion on the journey to 0'Kare Airport will not improve; in all probability it will get much worse. There is also a definite need to shorten the rush hour ground travel time to O'Kare. For these reasons, a rail link from the Loop ^ to 0 'Flare, free from the enormous delays of rush hour traffic, appears to be needed and would seem quite justified. The modal split model which is developed in the next three chapters will, presumably, indicate whether a rail link to 0 'Hare Air- port from the Lood would be used. - 8 - < ( CHAPTER IV - PROPOSED MODEL AND DATA GATHERING One of the objectives of this work is to develop a model for estimating the nodal split between taxi and airport limousine for trips from the Loop to O'Hare Airport. The selec- tion of an analytical procedure, which is thought to yield the best results, will now be discussed. Model Theory In the past several years, an accepted methodology has been evolved for analyzing modal split. Three basic sets of parameters (variables) are analyzed to arrive at a proper conclusion. These three sets which seem quite 'logical are system characteristics (such as travel time, cost, and frequency), personal characteristics (such as income, sex, age, educational level), and terminal characteristics (such as land use and popu- lation) . Most modal split analyses are generally used to predict the amount of public transportation use age in a metropolitan area. Either of two types of models, which are used by most con- sulting firms such as the trip-end model or trip interchange model, which was developed by the Traffic Research Corporation, 20 21 22 '" ' * are used for predicting modal split for all trips in an urban area. The other principal type of model is one which predicts individual choice of mode, such as that, developed by Wo T-'-n p v ^ • i ci.L lie Jl » i These models generally use either of two analytical - 9 - ( f procedures, regression or discriminant analysis. From the work 24 of Quarmby, it was found that for the particular case analyzed of the journey to work, regression analysis and discriminant analysis yielded equivalent results. Tor the work that was un- dertaken, it was decided to use the quite simple multivariate linear regression analysis. ^D >^° t^< )^& The modal is therefore of the form: A ^ bi Bp + bg B2 . . .+ bn Bn where A is the dependent variable, the percent modal split for a mode, and Bp through Bn the independent variables with their respective coefficients bp, bn. For the particular case at hand, it was assumed that the air travelers are now a fairly homo- geneous group such that, it would not be necessary to use the per- sonal variables in this analysis. Furthermore, because of the nature of the trip under analysis, it would also not be necessary ] ^ to use the terminals variables. Thus, the only independent vari- ables used in the analysis were those describing the actual opera- tion of each of the two modes. Upon analysis of several noteabie works in the field of modal split ,29 j^0?31 four measures of each mode were assumed to be needed. They were the following: travel time, cost, frequency, and access distance. *It was felt by the author that a detailed explanation of re- gression analysis was unnecessary and can be found in any of the rioted references. 10 - < Data Sources A primary data source was to be the 1964 Airport Survey conducted by the city of Chicago,32 hereafter referred to as the survey. For reasons explained at the end of this chapter, some survey data became "unavailable". For this reason, the model that was developed is theoretical and not practical. The survey Tvould have provided the split by mode for various areas of the Chicago Central Business District, the Loop, which was considered to run from Halstead Street to Lake Michigan and from Chicago Avenue to Roosevelt Road (See Figure 1). Even though these modal split data became "unavailable", the same analysis zones were used. The survey divided the Loop into thirteen analysis' zones (See Figure 2). The data that had to be gathered by the author for this work were, therefore, the independent variable measures of each mode from each of the thirteen zones. In other words, the travel time, cost, frequency, and access distance for each mode from the Loop to 0 'Hare had to be determined for each analysis zone. Thus, the crux of this analysis is that the modal split percentages supplied by the survey will be regressed against the independent modal variables, (This is the work of the next chapter). At this point, a procedural decision that was made about the data collected aid the resulting model, has to be stated. It was decided that the author would consider only trips departing from the Loop. A thorough description of the - 11 - i FIGURE 1 Chicago Central Busin ess D i strict 12 - < FIGURE 2 Analysis Zones S3GWNYOWN 13 - I operation of each mode will now be given from which some in- sight into data collection might be gained. Limousine & Taxicab- Routes & Rates From the work of Beimborn, a thorough understand- ing of taxicab operation in the City of Chicago con be attain- ed. Rates for Chicago taxicabs are based on distance traveled and waiting time* and are determined by City Council. From the Municipal Cede of the City of Chicago, * the applicable rates were found to be the following: 40^ for the first fifth of a mile and 10 $ eaGh third of a mile thereafter and 10c/ for each two minutes of waiting time. limousine bus service to 0 'Hare Airport from the Loop 35 is provided by the Continental Air Transport Co.. The limousines start at the Conrad Hilton Hotel, proceed west to State Street, north to Monroe, and then east to Wabash for the stop at the Palmer House Hotel. The buses then proceed east on Monroe to Michigan Avenue and north to the Sheraton Hotel for the third and last stop in the Loop, They then proceed north to Ontario and finally west to the Kennedy Expressway. The buses run at 10 minute headways during the day, seven days a week. Baggage handling is provided by the driver and the fare is §2.00 from all stops in the Loop. Hares for both modes are 1969 fares since ^Waiting Time is defined es the time consumed whenever the cab is not in motion due to unavoidable delays at street inter- sections or in traffic. - 14 - < 1964 fares and schedules for the limousine were unavailable. The reasons for this will be dealt with in the light of the modal split percentages (that will be discussed later). From the above description of each node, it can be seen that in order to obtain the required data without undue expenditure, taxicab and' limousine operations have to be simu- lated by the author. In other words, without going to the ex- pense of taking a taxi and limousine from, each analysis zone of the Loop to CfEare, the author used a private auto to simu- late the operation of the two modes. Analysis Zones The next step is to describe the thirteen analysis zones. In order to collect the required data, one must assume that a point, the centroid, within each zone can be considered es the point from which all trips from that zone originate. The centroid for each zone was located after examining each zone since many zones contain parks or railroad yards. The thirteen centroids, street intersections, that were selected arc listed in Table I. Field Data Collect ion A brief explanation of the procedure used to deter- mine the characteristics for each mode will now be given. First, however, several important points and assumptions have to be stated. The travel times were measured between 10 A.M. and 11 A.M. and between 1 P.M. and 3 P.M.. People beyond walking 15 TABLE I - ZONAL CENTROIDS 20NS CLNTROJD LOCATION 1 Grand Ave. & Milwaukee Ave. 2 LaSalle St. ?c Ohio St. 3 Grand Ave. & Michigan Ave. 4 Lake Shore Dr. & Ontario St 5 Clinton St. & Randolph St. 6 LaSalle St. & Randolph St. 7 Wabash St. & Randolph St. 8 Clinton St. & Jackson Blvd. 9 LaSalle St. & Jackson Blvd. 10 Wabash St. & Jackson Blvd. 11 Clinton St. & Polk St. 12 Dearborn St. & Folk St. 13 Wabash St. & Folk St. - 16 - distance, about 3 blocks, of the limousine were considered to gain access to the limousine by taxicab. This would seem to be a fairly reasonaole assumption since the survey found that almost 10% of the people used some other mode than walk- ing to reach the limousine. The other assumption is that the taxicab would take the fastest route leaving the Loop to get onto the Kennedy Expressway. An driving times for each mode were measured in two parts » Ail trips from the Loop were measur- ed for time end distance to a common point on the Kennedy Ex- pressway, quite close to the Loop. Then separate runs were made from that point to O'Hare. The actual data gathering for each zone is described in Appendix A. A summary of the data for each zone is compiled in Table II. The reasonableness of the driving times were found to be quite good when compared 37 with the work of Lisco . The author along with Lisco, found the mid-day travel time from the Palmer House to 0 ?Hare to average about 34 minutes. A very important point which has not yet been stated is that the unit of measure of "V.m in the data were arbitrary units. This was simply because the scale of the map used to estimate access distances was of an akward nature, i.e., Jw equals approximately 0.054 miles. Therefore, it was decided to carry out all computations with the arbitrary units, where 1 unit s 0.054 miles. - 17 I ■TABLE II DATA BASE ZONE MODAL SPLIT LT AT TT TO F W 1 Lime .20 34.0 5.0 39.0 2.00 2,60 10.0 21.0 1 Taxi .80 25.0 0 25.0 5.90 5.90 2.0 2.0 2 Li mo .50 34,0 3.0 37.0 2.00 2.40 10,0 8,5 2 Taxi .50 24.0 0 24.0 5.80 5.80 1.5 2.0 3 Li mo- .70 34.0 2.0 36.0 2.00 2,25 10.0 4.0 3 Taxi .30 29.0 0 29.0 6.10 6.10 0,5 2.0 4 Li mo .60 34.0 2.0 36.0 2,00 2,50 10.0 9.0 4 Taxi .40 27.5 0 27.5 6.30 6.30 2.0 2.0 5 Limo .20 39.0 6.0 45.0 2,00 2.60 10,0 23.0 5 Taxi .80 24.0 0 24.0 5.70 5.70 2,0 2,0 6 Limo .50 39.0 4.0 43.0 2,00 2.50 10,0 12,0 6 Taxi .50 26.0 0 26.0 5.80 5.80 1.0 1,5 7 Limo .60 39.0 2.0 41.0 2.00 2,20 10,0 6,0 7 Taxi .40 23.0 0 28,0 6.10 6.10 1.0 1,5 8 Limo .25 39.0 6.0 45.0 2,00 2,60 10,0 19.0 8 Taxi ♦ 75 26.5 0 26.5 6.10 6.10 1.5 2.0 9 Limo .55 39.0 3.0 42.0 2.00 2,50 10,0 10,0 9 Taxi .45 27.0 0 27.0 6.10 6.10 1,0 1.5 10 Limo .75 39.0 2.0 41.0 2.00 2 .13 10,0 3.0 10 Taxi .25 28.0 0 28.0 6.40 6.40 1.0 1.5 11 Limo .15 54.0 5,0 59,0 2.00 2,50 10,0 15.0 11 Taxi .85 28.0 0 28,0 6,20 6,20 2,0 2,0 12 Limo .30 54.0 3.0 57.0 2,00 2,40 10,0 7.0 12 Taxi .70 27.0 0 27.0 6,20 6.20 1.5 2.0 13 Limo ,50 54.0 2.0 56.0 2,00 2.10 10,0 3.0 13 Taxi .50 28.0 0 28.0 6.40 6.40 1,0 1.5 cXx. ± ablesr TT = = Total Travel Time TO = Total C ost LT = = Line Haul Time p = Headway AT = = Access Time 0 = Line Haul Cost w = Access : Distanci 3 18 - ; It should be noted that the modal split percentages. for each zone were also given In Table II. These percentages were to cone from the survey, however, due to political reasons beyond the control or comprehension of the 'author, it was de- cided after a lengthy wait, by an august assistant deputy com- missioner of Development and Planning for the City of Chicago that the data were "unavailable ."** It was then decided to synthesize the modal split data. Since it was known from the survey that of all the people leaving the=Loop by the two modes, fifty-five per cent used the limousine, it was assumed that a reasonable estimate, say within 10$ could be made as to the modal split percentages for each zone. The resulting modal split percentages were shown in Table II. The implications of this synthesis is that the model developed in the next chapter is a theoretical model and not a practical one. Because some 'assumptions and notable points concern- ing the data and, therefore the resulting modal, were many and were mentioned at different points in this chapter, it would seem most helpful If they were reiterated. They are the following :':<*The survey data is stored at the Chicago Area Transportation Study (CATS) data banks. After several phone conversations with the aforementioned deputy commissioner it was decided that the author could have access to the data. After the modal character- istics data were collected, Table II, it was decided by the di- rector of CATS that he would require more than verbal permission to release the data. After a lenghty wait of about two months, the decision was made that a letter releasing the data simply could not be written. At this point, the author was in what is known as a predicament. - 19 - f 1) Only system variables are used, 2) The data is for non-rush hour departures ♦ 3) The centroids can be used as point estimators of the zone, 4) People beyond walking distance of the limousine use taxicab to reach the limousine. 5) The taxicab routes from each zone are assumed the fastest. 6) 1969 taxicab and limousine fares and schedules are used. 7) The modal split percentages for each zone are based upon educated estimates. r 20 ♦ CIIAPTKR V ___- ...MODiJlL, .DW^^OH-g^jT _AITD .ANALYSIS The requirements of any mathematical model which at- tempts to explain human behavior such as travel are: 1) Accuracy - model can replicate base year conditions within acceptable tolerance ranges. 2) Validity - All independent variables are logical, both by their inclusion and coefficient. 3) Sensitivity - The model is responsive to any changes in the explanatory variables. Upon examination of the literature38,59,40,41,42} it was decided to examine four types of models in< the analysis. They are, for each independent variable, the following: 1) The ratio of each mode to the average for both modes (limousine - taxi ' ) (limo & taxi x limo & taxi) I 2 ■ """ 2 ). 2) The ratio of each mode to a hypothetically optimum mode ( limousine taxi ) (optimum optimum ). 3) The ratio to each other (limousine / taxi). 4) The difference between each mode (limousine ~ taxi). To insure immediate understanding, for example, in case four, the total travel time for the taxicab would be subtracted from the total travel time for the limousine, et cetera. The first stage in this model development was the selection of the best type of model. (Subsequently, the particu- lar model was further modified.) 21 - The first step was the analysis of the "ratio to bhe average" model (model type #1, above). Four independent vari- ables were used - the ratio to the average of the nodes of the total travel time (TT), cost (C), headway (F), and access dis- tance (W), Thus, there were 26 data points, two for each zone representing each mode. Because of the large amount of data to be considered in the model development, the BI.ID02R statistical computer program"-3 was employed in the multivariate linear re- gression analysis. This program calculates the mathematical relationships between variables by the stepwise approach. In the stepwise regression, all variables are input to the com- puter and correlation coefficients for all combinations of variables are calculated. The independent variable that ex- plains the greatest amount of variation in the dependent vari- able is selected for use in the model. During successive iterations or "steps," other independent variables are included in the model in light of previously entered variables and the criteria of explained variation. At each step the computer prints out the important statistical measures and the co- efficients of the equation. Por this case the resulting equation was M.S. (modal split) s*0. 481-0. 503TT -i-0.334C +0.743F-0.555W. The summary of this equation can be found with respective statistical computations in Table III, Stage #1. To determine 22 ""!5fc. TABLE III Stage //l Ratio to Average Ind, variables: TT*, C*, F, and W R - .909 foSt.Err. = 17.6 F-ratio ™ 2 5.08 Fs = 5.09 Stage jjz Ratio to Hypothetical Ind. Variables: TT*, 0*, F, antl W R = .768 $St.Err. = 27.3 F-ratio = 7.55 Fs = 5.09 Stage #3 Ratio to Each Other (Limo/Taxi)* Ind. Variables: TT*, C*, F*, and W* R = .3931 fbSt.Err. = 23.8 F-ratio = 7.88 Fq = 8.81 Stage //4 Differences (Lirao-Taxi) Ind. Variables: TT, TO*, F and W R a .969 v'st.Srr. » 13.9 F-ratio = 30.49 Fs ~ 8,81 Stage #5 Differences (F Dropped} Ind. Variables: TTj TC*, and W R a .961 $St.Err. = 14.5 F-ratio - 36.57 F3 b 8.72 Stage #6 Differences (TT replaced by IT & AT) Ind. Variables: IT, AT*, TO*, and W R - .961 ^St.Err. = 15.3 F-ratio = 24.44 Fs - 3.31 Stage #7 Differences (AT & TO Dropped) Ind. Variables: LT and W R - .957 ^St.Err. = 14.2 F-ratio = 56.76 Fq = 9.43 Stage §Q Final Equation (TT and W) SBM = ,867 - .013TT - .022V/ R = .959 f.St.Err. ~ 14.1 F-ratio = 57,39 Fs = 9.43 Legend * Statistically Insignificant at .01 Level of Significance F-ratio " Computed F Fs - Tabulated F Variables: TT - Total Travel Time TO = Total Cost LT - Line Haul Time F = Headway AT - Access Time C - Line Haul Cost -V.T a Access Distance 23 - ' whether the equation had a statistically significant fit, the F-test was performed on the equation. The F~test compares the variability about the fitted independent variables (numerator in F-ratio) with the inherent variability of the dependent vari- able (the denominator). A computed F-ratio value greater than that tabulated for the given conditions (degrees of freedom and level of significance) indicates a significant fit. The model was tested at the .01 significance level and was found to be highly significant. Using the partial F-ratio values computed .for the independent variables, the total travel time ratio (TT) and the cost ratio (C) were found to be insignificant as indi- cated by the asterisks in Table III. The second model type analyzed used the ratios of the modal characteristics to the characteristics of a hypothetically optimum mode. This was based upon the assumption that individual trip makers select their travel mode based upon a comparison to a hypothetical mode which would have apparent optimum characteris- tics. These were hypothesized as being a total travel time of 20 minutes, a cost of $1.00, a headway of 2 minutes, and an access distance of 3. There were thus 26 data points, as in stage §1, The B1.1D02R computer program was again employed and the following equation resulted: M.S. ~ -0.261 -0.205 TT + 0.164C + 0.21-1 F - 0.096V/ where the summary is in Table III, Stage //2. it should be noted 24 - ■that the nodal was barely significant by the F~test and the total travel time ratio and cost ratio were insignificant. The third Stage in the first portion of the analysis was the development of the model for the ratio of the variables, he,, the ratio of the limousine variables to the taxicab vari- ables. The BMD02R program was used again to perform the analy- sis with the resulting equation being: U.S. (Limousine) = 1.643 - 0.196TT - 2.298C + 0.006F - 0.032W where the summary can be found in Table 111, Stage #3. It is to be noted that the model was found to be insignificant. The final model type examined, uses the difference between characteristics of the two modes, i.e., limousine- taxicab. This model was also developed with the aid of the EMD02R program. The resulting equation was: U.S. (Limousine) ~ -0.079 -0.129TT - 0.078TC + 0.069? - 0.01517 where the summary is in Table III, Stage //4. The model was sig- nificant, although the Total Cost variable was insignificant. At this point, the first part of the analysis had been completed and all model types had been developed. It is obvious from an examination and comparison of the statistical summaries of each model type in the first four Stages of Table III that the use of "difference variables'' (stage #4) resulted in a su- perior model. It had the highest multiple R and F-ratio and the lowest % standard error. 25 - 01 The final part of this model development was concern- ed with the further modification of the "difference" model. As the first step in this part, the validity, or reasonableness, of the present model was examined. All the independent variables were logical and their coefficients were ail logical except for the difference in headways variable. This variable had a positive coefficient which meant that as the difference between the head- ways of the limousine and taxi increased the limousine would have received a higher share of the riders. This was highly illogical. Though, the reason for this result was quite logical. It so hap- pens that the zones with the highest use of the limousine also had the highest level of taxi service, or the greatest differ- ence in headways, since the limousine headways were constant. From the simple correlation matrix for the model, it was also found that there was a very high correlation 0.7, between the F and W variables. For these reasons and on the assumption that any new rail limousine service to the airport would main- tain the present schedules, the decision was made to drop the difference in headways^ variable , TfF" , from the analysis. The model was then reformulated with the "F" variable being dropped. The resulting equation was: M.S. (Limousine) ~ 0.612 - 0.013TT - 0.06 5TC - 0.019V/ where the summary is in Table III, Stage #5. There is little change in the' R and fo standard error as compared with Stage $4 but there is a noticeable increase - 26 - •II in the F-ratio. The total coat variable is again insignificant. As an attempt at further improvement, the total travel time variable was replaced by the line haul tine and the access tine variables. The following equation resulted: M.S. (Limousine) = +0,594 - 0.014LT - 0.007AT - 0.068TC - 0.02017, where the Summary can be found in Table ill, stage §6. When this equation was compared to the one developed in Stage $5, there was an increase in the percent standard error and a significant drop in the F-ratio. The access time and total cost were statisti- cally insignificant. The decision was then made to drop the variables which had been shown to be insignificant. An analysis was then per- formed just using the line haul time variable (LT) and the access distance variable (W) . The following model resulted: M.S. (Limousine) = 0.853 - 0.013LT - 0.024 W, where the summary is in Table III, Stage £7. Dropping the in- significant variables had caused the F-ratio to more than double compared with the previous analysis stage and had lowered the percent standard error. It is apparent that this was a much better equation. The only model which had not been analyzed with the insignificant variables dropped was the one with the total travel time and the access distance. The model resulting from this analysis was the following: M.S. (Limousine) = 0.867 - 0.013TT - 0.028W, - 27 - I where the summary for the equation is found in Table III, Stage #8. Upon, comparing the models developed in Stages 7 & 8, there was very little difference between the models. The total travel time model, Stage //8, however, had a slightly higher multiple R and F-ratio and a lower percent standard error. There fore, it was decided that the model developed in stage -''3 was the final and best model. The model expressing modal split as a percentage was simply: fo U.S. (Limousine) - 86.7 - 1.3TT - 2.2V/ where the "Multiple R" was 0.9591, the "percent standard error" was 14.1$, the "F-ratio" was 57.339 compared with a tabulated F of 9.43, at the .01 significance level and the "R-squared" was 0.9199. The correlation matrix for the model was the following: LiS TT W MS 1.000 -.637 -.831 TT 1.000 .201 W 1.000 as developed by the computer program. The access distance dif- ference variable, W, exhibited a strong negative linear correl- ation with the dependent variable while TT exhibited a lesser correlation.* Also, the colinearity between the two independent 5k .000 means no correlation. 1.000 or -1.000 means perfect linear correlation. - 28 - in variables was low, which is desirable for good regression re- sults. An important point which was stated previously was that the unit of measure of W in the model was an arbitrary unit, where 1 unit = 0.054 miles. The actual model, then, with travel tine differences in. minutes and access distance differ- enc e s in m i 1 e s , bee ame fo M.S. (Limousine) = 86.7 ~ 1.3TT ~ 43.717, however for ease of further analysis, the original form of the model is used in the next two chapters. .- 29 - ' CHAPTER VI ~ MODEL RESULTS As stated in the previous chapter for any model which attempts to explain human travel behavior such as modal split, 9 it has to meet three requirements, i.e., accuracy, validity and sensitivity. Therefore, the model, M.S. (Limousine) = 0,867 - 0.013TT - 0.022W, must meet each of these ■ requirements to be acceptable. For the model to be accurate, it must replicate the base conditions within acceptable tolerance ranges. The r?j for this equation was 0.9199, which meant that the model explains some 92$ of the variation in the data. The percent standard error was slightly high at 14.1$, but this was still quite ac- ceptable. A comparison of the computed IP-ratio to the tabulat- ed F, 57.4 to 9.43, indicated that the model was highly statis- tically significant. In summary then, the model quite accurate- ly replicated base conditions as portrayed by the data. There is an important point, however, which can not be overlooked, this Is that the percentage split by mode for each zone, the dependent variable, was based upon an "educated estimate." The question that now has to be answered is what effect would a change in the assumed modal splits for each zone have on the model. Since it is believed that the estimates of the fj modal splits were off by not more than about ten percentage 30 - points for each zone, as stated previously,* an analysis of the effect of a small change in the estimated nodal split percent- ages on the model was made. To accomplish this, small random Qhanges were made in the dependent variable data and the regression equations resulting from the random changes were compared and analyzed. Each dependent variable data point was changed from zero to plus or minus twenty percentage points in a random fashion, in order to accomplish these random changes, a random number table was used. The following procedure was employed: 1) the first zone was assigned a random number, 2) the rest of the zones were assigned the numbers from the random number table which follow the first number already assigned, 3) the limousine $ modal split for all zones with random numbers between zero and two were changed by plus 20$, between two .and four by plus 10$, between four and six by zero $, between six and eight by minus 10$, and between eight and ten by minus 20$. The resulting equations from the above procedure and regression analysis are shown in Table 17. Each step with the same number represents the same random number series. A letter indicates * It must also be remembered that the weighted average modal split for the entire Loop is 55/45, limousine/taxicab. 31 [I I TABLE IV EQUATIONS RESULTING PROM RANDOM ORANGE Equation Change Constant .867 TT Ooeff . -.013 W Coeff . -.022 P- Lev el Ra | 57.4 .959 % St.Er: #1 None -*- 14.1 .#2 0 ■I- 5 5 10 10 .919 -.013 -.029 55*3 .957 18.0 #3' + 10 + 5 0 5 10 .894 -.013 -.025 53.1 .956 16.3 #4 + 10 + 5 0 5 10 .844 -.010 -.022 13.2 .852 27.0 #5a + 20 + 10 0 10 20 .772 -•010 -.020 7,4 .772 38.2 #5b + 10 + 5 0 5 10 .819 -.012 -.021 21,1 .900 22.8 #5o + 5 + 3 0 3 5 .840 -.012 -.021 36.6 .938 17,4 #6a + 10 + 5 0 5 10 .815 -♦010 -.022 18.7 .888 25.0 #6b + 20 + 10 0 10 20 .739 -.008 -O023 6.0 .739 44.6 #7a + 10 + 5 0 5 10 »960 -.014 -.028 77.5 .970 14.2 #7b + 20 + 10 0 10 20 1.05 -.015 -.034 40.9 .944 22.0 #'8a + 10 + 5 0 5 10 .751 -s006 -♦021 12.7 ,847 24.2 #8b + 20 + 10 0 10 20 .663 ~ . 002' -.020 4.3 .680 48,0 #9a + 10 + 5 0 5 10 .957 -.017 -.022 31.8 .930 21.5 #9b 20 + 10 0 10 20 1,05 -.022 -.023 17.6 .882 32.7 32 ' that the sane random number series was used with different per- centage changes. For example, $9a was changed from +10$ to -10$, but #9b was changed from +20$ to -20$, however, both //9a and #9b used the sane random number series. The results of this analysis indicated that while changes in the dependent variable data of up to 10$ caused little change in the model, larger changes, that is, of up to 20$ started to alter the model. Therefore, even if the estimated modal splits for each zone were slightly off, the model would be substantially the same. It is now felt that the model does fulfill its accuracy requirements. It must still be re- membered though, that the model is only a theoretical one. The next point to consider is whether the model is valid, or reasonable. The mode choice being attributed to only the difference in travel time and difference in access distance between the limousine and taxicab seems entirely logical and plausible. The negative coefficients are logical, obviously since as the difference between the travel times and access d i s t an c e s In c r e a s e s , the II mou s in e w o ul d lose patrons. One point, however, needs some discussion, that is, the insignificance of the cost variable. This at first, seems illogical, but when one considers that in all probability the people making the modal choice are businessmen on expense ac- counts, then the insignificance of the cost in the decision making process begins to appear more plausible. A very important consideration concerning these last 53 - < two points, however, must be discussed* That is the possibility that the cost was shown to be insignificant in the analysis primarily because it was not properly considered in the estima- tion of the zonal node splits. No matter how hard the author attempted to estimate the nodal splits for each zone on an un- biased basis, there is always the possibility that the effect of the cost on the mode choice decision was not fully considered. In retrospect it is felt by the author that the full effect of cost was not completely considered - the major basis on which the estimation procedure was carried out being the distance one must travel to reach the limousine. This is why, conceptually, the model could have possibly ended up in its present final form. On the other hand, other mode choice studies were examined to see if their results concur with the apparent result of this work as to the seeming insignificance of cost. Quarmby's study . indicates that cost is not as significant as various other fact- ors affecting the .node choice for the journey to work. An 0~D study of railroad commuters conducted for the Southeastern 45 Pennsylvania Transportation Authority" concluded that fare did not seem to have as much influence as is sometimes ascribed to it. (Another interesting conclusion is that very few traveled more than 10 minutes by foot or transit from their CBD arrival point.) A study conducted by the Charles River Associates finds that free transit service in Boston would not stimulate patronage as much as improved transit service would. Two public attitude - 34 - . surveys toward transit, one conducted by the University of 47 Maryland and the other conducted for the Twin Cities Area Metropolitan Transit Commission , finds cost not to be as im- portant as such factors as travel time, reliability, comfort and safety. Thus, the absence of cost from the model seems to have some plausibility over and above possible biases intro- duced in the data. The last requirement of the model was that it be sensitive to chenges in the independent variable^. To determine if this requirement was met, a sensitivity analysis was perform- ed on the model. The relative sensitivities among the independent vari- ables of the model were gauged by computing the "absolute spread 0f each variable and by comparing it with that for other variables in the equation — , the larger the spread the greater the sensitivity of the dependent to the independent variable. "Absolute spread" was calculated by multiplying the limits of one standard deviation about the mean (mean plus stand- ard deviation & mean minus standard deviation) of each variable by its appropriate regression coefficient and then by determin- ing the difference between these limits. The application of this is illustrated in Table V, Sensitivity Analysis. To further illustrate the varying degrees of sensi- tivity, the two independent variables were each, separately, Increased 50$ beyond their individual mean values, while the - 35 - SENSITIVITY ANALYSIS 1.1.3. % (Limousine) = 86.7 - 1.3TT - 2.2V/ Mean Standard Deviation M . S.fo 44 . 6 20.3 TT W 17.615 9.076 7.013 6.833 Weight in Equation TT High (17. 615+7. 813)(1.3)" 32.0 Low (17,615-7.813) (1.3)= 12.0 W ( 9 . 076+6 .833 ) ( 2 .2 ) »35 . 0 (9,076~6.833)(2.2)~ 4.0 Absolute Spread 20.0 31.0 Reaction to Adjustments in Independent Variables Condition All Variables © Mean 50f^ Increase in TT 50% Increase in W C of -'IP u t e d Fere en t 44.6 33.1 35.1 Gliuinge No, Percent -11. 5 9 R 5.3 - 9.5 -21.1 - 36 - other variable was held cons bant at its mean value. The comput- ed value of the dependent variable was compared with its value at mean conditions. The results of this can also be found in Table y. It can be seen that a 50$ increase in the independent variables above their means caused a change of from ZVfo to 2G% in the dependent variable. Obviously, the model is sensitive to changes in its explanatory variables. At this point, one of the objectives of this study has been achieved, that is, the development of a model for predicting modal split between airport limousine end taxicab for trips from Chicago* s Loop to O'Hare International Airport.. The model ex- plains the use of the limousine as a function of the difference in total travel times and the difference in access distances between the limousine and taxicab. Now one can consider what the implications of this model are for a possible future rail limousine service to O'Hare Airport. - 37 ( IMPLICATIONS FOR A POSSIBLE PUTURE HIGH SPEED RAIL LII-IpUSINE SERVICE TO O'KARE AIRPORT 0 The point of Chapter III was that present ground ac- cess to O'Hare, entirely by highway, simply cannot meet the needs of air travelers by the end of this decade. At this tine, two high speed rail links to O'Hare from the Loop have been discussed as a way to help relieve the ground access problem. One alternative is to simply extend the newly opened Chicago Transit Authority "Kennedy" rapid transit line from Jefferson Park in the median of the J.P.Kennedy Expressway to O'Hare. The other alternative is to run Chicago & Northwestern Railroad trains out along the median of the J. F. Kennedy Express- way from Jefferson Park as the abovementioned CTA. alternative. (The CTA Kennedy line now terminates at the Jefferson Park sta- tion of the C.fc N.W.R.'R.) Prom the nodal split analysis which preceded, it is quite logical to ask the question as to what the implications are of the nodal split model, for evaluating the above rail access alternatives. What is to follow is a theoretical exercise to determine if there is any substantial difference in the attract- ing power of either rail alternative, remembering, of course, that the model is only theoretical. Before such an analysis can be performed however an 38 - 1 V * assumption, which is quite plausible, has to be made, Either of two rail alternatives, it is assumed, will have all the compar- able amenities of the limousine service such as seats for all, airconditioned vehicles, and baggage handling. A description of the physical facilities and attri- butes of each alternative will now be discussed. The railroad alternative would terminate in the Loop at the present North- western Station at Clinton and Randolph Streets. It would operate with present C .& N.W. equipment, bi-level push-pull trains, over the C.& N.W. right-of-way from the Loop to Jefferson Park, on its northwest line, and then proceed out along an extension in the median of the Kennedy Expressway to O'Eare. Running time, based on present schedules for a similar distance, would be approximately 23 minutes, or an overall average speed of 45 mph. Service would be provided at the same headways of the present limousine service, ten minutes. The rail rapid transit alterna- tive would be an express service from the Loop to O'Hare over the present GTA subway and elevated right-of-ways. Assuming the loop subway is constructed, a special subway airport terminal could be built at Monroe and Wabash, where the present limousine terminal is located. Trains would operate around the loop subway, then out along the Kennedy line to Jefferson Park where an exten- sion would be built continuing, in the median of the Kennedy Expressway to O'Hare. Running time is estimated at 30 minutes t 39 - < III assuming a 23-i- minute non-rush hour running time to Jefferson 50 Park,' because the airport trains would have to follow regular CTA trains. Headways would also be ten minutes. It must be stated that possible operational delays were not considered in estimating the running times for each rail alternative. The physical aspects of each alternative having been described, a modal split analysis for each alternative was made. Only the major traffic generating analysis zones are considered in this analysis. These zones are nos . 2,3,5,6,7,8,9 and 10 t (See Figures 2 and Table I.) It was assumed that taxicab travel times would not change from the present. The average difference in travel times and access distances between limousine and taxi- cab were determined for each rail alternative. The average of these times and distances are shown below with the present limousine included. C.& N.W. CTA Limousine Alternative Alternative f present) Travel Time Difference 0.3 minutes 6.9 minutes 19.1 minutes Access Distance Difference 0.64 miles 0.60 miles 0.47 miles Thus, the actual difference between the two alternatives is 6.6 minutes, (6.9-0,3= 6.6) and 0.04 miles (.64-. 60 ■ .04). The C.&N.W. alternative is 6.6 minutes faster, while the CTA alternative is 0.04 miles shorter. Using the model . [fo M.S. (Limousine) = 86.7 - 1.3(TT) - 43.7 (W) ) 40 - ' .' and the results of the preceding table, the nodal split percent for the C.&N.W. alternative is 58.5$, while the mode choice per- !. cent for the CTA alternative is 51.9$. The present limousine service draws 48.5$. The difference between the two alternatives y is 6,6$ with the C •& N.Y.'.R.R. ' alternative being superior. From the above analysis, a high speed rail limousine service to O'Hare leaving from Northwestern Station and using C.& N*W. railroad equipment and right-of-way would therefore attract 7$ (nodal split $) more riders from the major traffic generating areas of the Loop in non-rush hours then, an altern- ative using the CTA right-of-way* It should be _ remembered that the major assumptions used to arrive at the above conclusion were that rail limousine service included the proper amenities and that each alternative would have only one terminal in the loop. If the CTA service alternative were to stop at several other points in the Loop for those Without baggage, this obvious- ly would improve its attractiveness. An additional stop at LaSalle and Randolph handling baggage would entirely eliminate the C.& N.W. alternative's advantage due to the reduced access distances. It can be seen that the "model" is useful in evaluat- ing the demand side to both rail limousine alternatives. The determination of the best alternative can only occur after a detailed engineering cost study of construction costs has been - 41 - conducted, and when a practical node split model, not a theoreti- cal model, has been developed. Then, the complementary supply (or cost) and demand sides of each alternative could be analyzed together to determine which is the best alternative. - 42 - I GRAFTER VIII CONCLUSIONS A theoretical model has been d ev eloped that will pre- dict the modal split between airport limousi ne and taxi cab for trips from the Chicago Loop to O'Hare Airport as a function of the difference between the two modes in travel time and access distance. The differences model was found to be superior to a model using the ratio between the modes as variables. The final model was the following: at 7° M.S. (Limousine) = 86.7 - 1.3TT ~ 43. 7W, where TT and VT are measured in minutes and miles respectively. The equation is logical, reasonable, and sensitive. This state- ment can be made principally because the random change analysis successfully demonstrated the low sensitivity of the independ- ent variables to the possible errors in the dependent variable estimates. Considering the assumptions that were made, the model responded as expected. It was found from the analysis that the cost was insignificant in this quite interesting mode choice decision. However this cannot be stated with certainty due to the possibility of a biased zonal mode choice estimation pro- cedure, even though this confirms what other investigators have discovered about cost and its position in affecting mode choice. That is, that cost is not as significant a s previously surmised. Furthermore, this was apparently concluded in an area never before investigated. Thus, the model performed similar to - 43 - I I) ( more standard modal split models. One outcome of this model is that it can be applied to improve the market position of the present airport limousine service by simply adding a west Loop departure terminal. The model was quite useful in evaluating the demand aspects of possible future rail limousine alternatives. The rail limousine line using the C.& N.W.R.R. right-of-way instead of the C.T.A. right-of-way from the Loop to Jefferson Park has the superior market position, based on only one Loop Terminal. It should be noted that either rail alternative is not very much better than the present limousine in non-rush hours. This is only the first step. It represents one of the first times that such an interesting modal split model has ever been developed. Tnis can be the starting point. As the first step in extending this analysis, a study could be undertaken with complete data to develop a practical model and, also, since this theoretical model was only constructed for non-rush hours, develop a similar rush hour model. A similar study should be made in other cities to see if one model is applicable only in Chicago or country-wide. Another quite logical extension is the analysis of non-linear regression equations and the investiga- tion of variables that might have been overlooked. With this analysis as a stepping stone, "the sky is the limit." 44 - I ( 11. 12. 13. 14. 15. BIBLIOGRAPHY 1. Popper, R., et al . Airport Transportation ~ A study of .Tran sportati on Mean's^ etween~~A 3, rporfc s lTncT"the*"* MetrojDO_litan Xreas they serve. Human Sciences Research, Inc., Arlington, Virginia, February, 1961. 2. Cleary, E. J. and Whitlock, E. M. "Planning Ground Transportation Facilities for Airports." Highway Re search Record #274, Wa shingt on , D . G . , T§&$ . 3. Popper, R., et al, ojw cit_._ 4. Ibid. 5. Silence, Steiner M. A Preliminary Look at Ground Access to Airports. Highway Research Record #274, Washington, D. C77~TJ61T. 6. Schriever, Bernard A. Air Tran sportation 1975 and Beyond; iLJ&lzP iSL-^£E£^25I Cambridge, Massachusetts^ 19bT7 7. Department of Transportation, Federal Aviation Administra- "ki°n. ^ZJ^M^L ■F.ore(ias-^ 1969-1980 . Washington, D. C, January, 1969. 8. Lockheed-California Corporation. Air Travel Demand 1967-1990. 1967. ____ 9. Schriever, Bernard A., op. cit. 10. Tardiere, Salvatore G. and Jarema, Frank E. "Impact of Projected Air Travel Demand on Airport Access." Highway Research Record #274. Washington, D. C, Gatewich Airport. Going Places^ Second Quarter, 1968. Popper, R., et al, op. cit. Possible Solutions to Improve Airport Access, Aviation Weekly. Vol. 86 #18. May 1967. Fly Now with Cleveland Transit. Passenger Transport. Vol. 26 #32. November 157~196=8^ ~ "Kennedy Airport Rail Link Approved by New York Board." Passenger Transport. Vol. 27 #33. Washington, D~7~C". , November, 1969. - 45 - « ( 16. "London's Airport to Use Rail Rapid Links with City." Pass eng er Tr an s p or t . Vol. 27 #26. Washington", D. C, October I9~6~9. 17. Lisco, Thomas E. "Airport Access." Driving Times between O'Hare Airport and Downtown Chicago, Chicago Area Transportation Study Research News. September - October. T9Eq;~~ — 18. Tardiere, Salvatore G. and Jarema, Frank E., op, cit. 19. City of Chicago, Department of Development and Planning. O'Hare Airport Passenger Survey. Ho v ember, 1965. 20. Hill, D. M. and Von Cube, H. G., "Development of a Model for Forecasting Travel Mode Choice in Urban Areas," Highv;ay__Research Record. #38, Washington, D. C, 1963. ^* Some A spect s of Pi scriminant Functions and Other Interurban M o 6 a 1 Split Mo cTeTs . Traffic PTe sea rch Corporation to N.E.C.T.P. National Bureau of Standards, Washington, D. C, 1965. 22. Sosslan, A. B., Heanue, K. E., and Balek, A. J., "Evaluation of a Hew Modal Split Procedure" Highway Research Record. #88, Washington, D. 0., 1965. — ™~~ *"~ 23. Warner, S. L., Stockastic Choice of Mode in Urban Travel. A Study in Binary Choice^ Northwestern University, Evan st on , I ll Inois, 1962. 24. Quarmby, D. A,, "Choice of Travel Mode for the Journey to Work," Journal of Transport Economics and Policy London, September, 15J5T7 " ™ """" " 25. Greenshields, Bruce D. and Weida, P. M., Statistics with Ajpplicati on s to Highway Traffic Analyses*. * E"N~~0 Foundation, 1552. 26. Miller, I. and Freund, J. E. , Probability and Statistics for Engineers. Prentice-Hall, Inc., 1955. 27. Guttman, I. and Wilk, S. A., Introductory Engineering Statistics, John Wiley & Sons, Inc., New YorFj IU65. 28. Williams, E. J#, Regression Analysis , John Wiley & Sons, Inc., New York, 19 5!5T 29. National Bureau of Standards, op. cit. -. 46 - i 30. Warner, S. L., op. cit. 31. Quarmby, D. A., op. cit. 32. City of Chicago, op. cit. 33. Beimborn, Edward A., "Characteristics of Taxi cab Usage." Masters Thesis, Northwestern Univer si t.y , Evans ton, Illinois, June l§5b . 34. Municipal Code of the City of Chicago. Section 23-30. June, "1969. 35. Information obtained from Continental Air Transport Company, a Supervisor and Schedules. 36. City of Chicago, £P^__2iii 37. Lisco, Thomas E., op. cit. 38. National Bureau of Standards, op. cit. 39. Warner, S. L., op. cit. 40 . Quarmby , D . A . , op. cit. 41. McLynn, J. M. and Watkins, R. H., "Multimodal Assignment Model," Davison, Talbird, & McLynn, Inc., 1965. 42. Baumol, W. J. and Quardt, R. E., "The Demand for A.bstract Transport Modes: Theory and Measurement," Journal of Reginal Sciences. Vol. 6 #2, 1966. 43 • Health Sciences Computing Facility, Stepwise Regression, ~~~~BMD~T)'2R, U C L A, December, 1965. 4 4 . Quarmby , D . A . , op. cit. 45. Southeastern Pennsylvania Transportation Authority, Ridership_ surveys of Pennsylvania Railroad commuter II ne s^Ji^YiJ3 1 0imj_P ^l^Ji1^1 ^ Hi 11 , Manayunk and" Pa6li_ Lilies and all 'Reading Railroad' commuter trains "operaTing In Southeastern Pennsylvania , Philad™eTphTa, N 6 v"emb er , 1 $E'6~, 46. Charles River Associates, Evaluation of Free Transit Service r Boston, 1968. 47. Fash, A. N. and Hills, 5. J., "Public Attitudes Towards Transport Modes," Highway Researcn Record #233, June, 1968. . r 47 - « » 48. Simpson & Curtin, Public Attitudes Toward Transit, Philadelphia, April , T9o§ . 49. Corradino, J. C, "The Effect of the Highway System and Land Development on Trip Production," Traffic Engineering, June, 1968, 50. From private correspondence with Mr. F, I-lisek, Asst. Chief Engineer of Research and Planning, Chicago Transit Authority. _ 48 '♦ I APPENDIX ZONAL FIELD DATA COLLECTION A brief explanation or some special aspect will now be made of each variable appearing in Table II. Then, each zone will be discussed separately explaning how all the vari- ables for each mode were obtained. The variable "LT" is the line haul time for each mode. It is the time spent in eacn mode on the trip to O'Kare. For the limousine, there is five minutes (two plus three) added to its line haul running time for the two following reasons. One is because it was observed that the limousine spent two minutes, on the average, at the Sheraton Hotel stop on North Michigan Avenue. The other is due to the limousine having to make five stops at O'Hare. It was observed that the limousine took ap- proximately tnree minutes more to reach the middle, or the third stop at O'Hare than a taxi to reach the same location. The "AT" variable is the travel time spent in gaining access to the limousine or taxi from one's boarding point. The access mode to the taxi is walking and was considered as zero. The access mode to the limousine was considered as taxi unless the distance was less than three blocks. The access time was simply the time spent walking or the time spent in the taxi. The "TTn variable is simply the sum of "LT" and "AT". The cost variables, "C" and "TC" are based on the rates of fare for each mode. The total cost,. "TC", includes - 49 - i the cost of the taxi to reach the limousine. The "F" variable is the headway at which the lirao- sines are scheduled and the time between empty taxis arriving at the zonal centroids as observed in the field. The "W" variable is the distance one must travel to gain access to the limousine or taxi. The distances were measured from a map with a scale of one inch equals approx. 0.216 miles. Due to the akward nature of this scale, all analytical computations were carried out using an anonymous unit of 1 unit = 0.054 miles (ir!,= 0.054 miles). The units are such that in the center of the Loop, one block equals roughly 2 units. Zone 1 Zone 1 lies in the Northwest corner of the Loop and is bounded by Halsted St., Chicago Ave., Orleans St., and an alley one block north of Lake St.. This area mostly consists of the North Branch of the Chicago River and the railroad yards of the C .& N.W. and Milwaukee Road. The only business activit3'' centers around Grand and Milwaukee Aves.. Limousine An individual would use the taxi to reach the limousine. The access time was 5 minutes by taxi which proceeded east on Grand to Michigan. The fare would be 60^. Thus, the total cost would be $2.60. The line haul time to O'Hare from the - 50 - < ( Zone 2 Sheraton was 34 minutes. Total travel time was 39 minutes. Taxi cab The taxi cab was assumed to proceed east on Grand , to Orleans, north to Ontario, and west to the Kennedy Expressway. The travel time was 25 minutes and the di stance was 18"; 5 miles. The fare would be '|5.90* Zone 2 is bounded by N. Orleans St., W. Chicago Ave., Ho. State St., and W. Waeker Dr. Limousine Taxi is again used to gain access to the limou- sine. The access time was 3 minutes. The taxi route was east on Ohio to Michigan and South to the Sheraton. The taxi fare was 40^. The total cost was $2.40. The travel 'time was 34 plus 3, or 37 minutes. Taxi cab The taxicab route was north on LaSalle to Ontario, and west to the Kennedy Expressway. The travel time was 24 minutes, and the distance was 13.2 miles. Tne fare would be $5.80. Taxis passed by about every 90 seconds. - 51 - Zone 3 Zone 4 Zone 3 is bounded by E. Chicago Ave,, a distance one block east of Fairbanks Court, Wacker Drive, and M*. State St., This zone contains the Sheraton Hotel. Limousine The zone is such that roughly five- eighths of the people require a taxi. As a result the access time is two minutes and the access cost in only .25^, the total cost being $2.25. The travel time was 36 minutes . Taxi cab The taxi uses the same route as the limousine from the Centroid of this zone. The travel time to C'Hare was 29 minutes, the distance was 13.7 miles, and the waiting time just exceeded two minutes. The resulting fare was -v'o.lO. Empty taxis passed by every 30 seconds. This zone lies along the Lake from Chicago Ave, to roughly the Chicago River. It includes the Navy Pier. Limousine Access was by taxi which took only 2 minutes to reach the Sheraton Hotel. Taxi fare was 50^ and the total cost was thus $'2.50. Total travel time was 36 minutes. - 52 - i Zone 5 Zone 6 Taxicab Taxis simply proceeded west on Ontario St. to the Kennedy Expressway. Travel time to O'Hare was 27.5 minutes, the distance was 19.3 miles, and the waiting time exceeded two minutes. The resulting fare was $6.30. Taxi service is not as good in this zone as along Michigan Ave,, and taxi cabs only passed by every two minutes. Zone 5 lies on the west side of the Loop and is bounded by Halsted, Madison, Hacker Dr., and an alley one block north of Lake St.. Limousine Taxi is used for access to the limousine at the Palmer^ House taking 6 minutes and costing 60/. The total cost is $2.60 and the total travel time is 45 minutes.. Taxicab The taxicabs gain easy access to the Kennedy by proceeding west on Randolph from Clinton. The travel time was 24 minutes, and the distance was 17.8 miles. The cost would be ^5. 70. The taxi frequency was observed as two minutes. This zone is bounded by Wacker Dr. on the north and west, ¥. Madison St., and N. State St.. 53 - Zone 7 Zone 8 Limousine Access was again by taxi which took 4 minutes and cost 50/. The total cost was (';2,50, and the total travel time was 43 minutes. Taxi cab The taxis were assumed to proceed west on Randolph from LaSalle to the Kennedy Expressway. The travel time was 26 minutes and the distance was 18.3 miles. The fare would be $5.80. Frequency was observed as one minute. This zone lies between Wacker Dr.-, IT. State St., E. Madison St., and Lake Michigan. Limousine Access is gained by both walking and taxi cab. The access time is- 2 minutes and cost is 20^. The travel time is 41 minutes and costs $2.20. Taxi cab Taxis are a ssumed to proceed west on Randolph from Wabash to the Kennedy Expressway. Travel time was 28 minutes, the distance was 18.8 miles, and the waiting time exceeded two minutes. v The fare would have been $6.10. Zone 8 is bounded by W. Madison St., S. Wacker Dr., W. Harrison St., and S. Halsted St.. .- 54 - Zone 9 Zone 10 Limousine Access was by taxi which would proceed east on Jackson Blvd. and north on Wabash, It took 6 minutes and cost 60^. The total time was 45 minutes and cost $2.60. Taxi cab Access to the Kennedy Expressway from Clinton and Jackson was north on Clinton, west on Adams, south on Hal s ted and east on Jackson. The trip to O'Hare took 26.5 minutes, covered 19 »1 miles, and cost $6.10. Taxis passed by about every 90 seconds. Zone 9 is bounded by W. Madison St., S, State St., ¥. Harrison St., and S. Wacker Dr.. Limousine Access was again by taxi to the Palmer House and took 3 minutes and cost 50/. The total cost was $2,50 and the trip to O'Hare took 42 minutes. Taxi cab Taxis from this zone were assumed to travel south to Congress St., west onto the Eisenhower Express- way and then north into the Kennedy Expressway. The travel time was 2 7 minutes and the distance was 19.5 miles. The fare would have been $6,10. This zone contains the Palmer House Hotel. The ~ 55 Zone 11 •- zone is bounded by Harrison, State, Madison, and Lake Michigan. Limousine Access is gained by walking and taxi. Access cost was 13/ and time was two minutes. The total time was 41 minutes, and the total cost was $2.13. Taxi cab The cabs were assumed to reach the Kennedy Express- x-jay by proceeding south on Wabash to Congress St. and then west. The travel time was 28 minutes and the distance was 19.8 miles. The fare would be $6.40. Frequency was observed as about one minute. This zone is bounded by Harrison St,, the Chicago River, Roosevelt Rd*, andHalsted St.. Limousine Access is by taxi to the Conrad Hilton. Access time was five minutes and cost 50/. The total time was 59 minutes and cost ^2.50. Taxi cab The assumed route of the taxis from Clinton and Polk was south on Clinton St., to Roosevelt Rd., and west to the Kennedy Expressway. Traveltime was 28 minutes, the trip covered 19.5 miles, and cost #6.20. .- 56 - Zone 12 Zone 13 Zone 12 is bounded by zone 11 on the west, W. Harrison, S. State St., and W. Roosevelt Rd . . Limousine Access was by taxi which took three minutes and cost 40/. The total cost was $2.40 and the trip took 57 minutes. Taxi cab The taxis would proceed north on Dearborn St., west on Congress St,-, and onto the Eisenhower Expressway, and then north on the Kennedy Ex- pressway. The trip to O'Hare took 27 minutes, covered 19.5 miles, and cost $6.20. This zone contains the Conrad Hilton Hotel and is bounded by E. Harrison St., S. State St., E. Roosevelt Rd., and Lake Michigan. Limousine Access was gained both by walking or taxi. Time was 2 minutes and cost 10/. The total travel time was 56 minutes and cost $2,10. Taxicab The access route to the Kennedy Expressway was north on Wabash and west on Congress St.. The trip to O'Hare took 28 minutes, covered 19.8 miles, and cost $6.40. t 57 -